Virtual agents, such as chatbots, emailbots, salesbots, and/or AI assistants, may be deployed with intelligent algorithms working in the background to assist users with inquiries and provide suggestions, directions, and/or other useful information. When users are interacting with the virtual agent, the virtual agent interprets and responds to natural language utterances. A virtual agent under test may be trained with various conversational scenarios to increase the accuracy and authenticity of the virtual agent. Present approaches to conversationally testing and training virtual agents involve manual effort to generate test data with limited test coverage.